Data Analytics for Medical Record Data of Covid-19 Patient with Descriptive & Predictive Mining

Currently, the number of COVID-19 patients in Indonesia has not shown a significant decline. One of the reasons is the difficulty of analyzing medical record data of COVID-19 patients. The analysis becomes difficult because of the large volume of data and the heterogeneous characteristics of the content. This is what makes medical personnel take a long time in the process of analyzing data on COVID-19 patients, and causes the slow handling of cases in Indonesia. Therefore, this study proposes a new approach to analyze and classify medical record data of COVID-19 patients using a data analytics approach with descriptive and predictive mining methods. Where descriptive method is used to grouping patient’s disease risk categorization and predictive method is used to classify positive/negative patients. The input data used for the analysis process was sourced from the results of the RT-PCR test of COVID-19 patients. Before being analyzed, the data is normalized first, and the features are extracted. Next, the descriptive mining process was carried out using the Hierarchical Clustering Complete Linkage method and predictive mining using the Decision Tree and K-Nearest Neighbor (KNN) methods. If so, then the results of the process are tested using the confusion matrix and error ratio methods. From the whole process, it can be concluded that the error ratio of this approach is 8% for predictive mining and 21% for descriptive mining. And the features found at the end of the process were 19 features, including Leukocytes, Eosinophils, Red blood cells, Patient age quantile, etc.